Of the dozens of loci genome-wide association studies have linked to Alzheimer’s, most are mysterious to scientists. How do they contribute to disease? Much work focuses on changes in expression of individual genes, but as reported in a preprint posted to medRxiv on February 8, scientists led by Celeste Karch and Oscar Harari at Washington University in St. Louis took a different approach. They used single-nuclei RNA-Seq data to dissect the effects of risk variants on whole human cell populations. In so doing, they linked a protective variant in the MS4A4A gene to expansion of an anti-inflammatory subtype of microglia that expressed interferon and lipid metabolism genes. Conversely, a risk variant in the MS4A4A gene suppressed this cell state, turning up pro-inflammatory cytokine pathways and hampering lipid metabolism. Karch told Alzforum she believes MS4A4A serves as a major regulator of microglial states.

  • A protective variant in the MS4A4A locus promotes anti-inflammatory microglia.
  • A risk variant in MS4A4A represses them.
  • These microglia express interferons and metabolize lipids well.

For Michal Schwartz at the Weizmann Institute of Science in Rehovot, Israel, the data help clarify how the MS4A locus contributes to AD and focus attention on lipid biology. “This study highlights the connection between cholesterol metabolism and inflammation, a known phenomenon in other cells and organs, but novel in the context of microglia,” Schwartz wrote. Julia TCW at Boston University was also impressed by this aspect. “This is interesting work, showing that interferon signaling induced by the protective microglia variant increases cellular cholesterol efflux. That may potentially reverse the cholesterol accumulation phenotype found in glia, one of the AD hallmarks,” she said (comments below).

MS4A4A, a Switch? In disease conditions, homeostatic microglia can transition to either a pro-inflammatory (Mic 1) or anti-inflammatory (Mic 3) state. Mic 3 cells are themselves influenced by protective (green) and risk (red) SNPs in the MS4A4A locus, which push these cells into either a helpful, interferon-based response with efficient lipid metabolism (left), or a destructive inflammatory response that accumulates lipids (right). [Courtesy of You et al., medRxiv.]

Variants at the MS4A locus were first linked to AD more than a decade ago (Apr 2011 news). This multi-gene locus encodes several microglial transmembrane receptors. With Carlos Cruchaga at WashU, Karch identified a functional SNP, rs1582763, in this region that boosts expression of one of these genes, the lipid sensor MS4A4A. The variant seems to stimulate TREM2 processing, correlating with elevated sTREM2 in CSF, and it lowers AD risk (Jul 2018 news; Apr 2019 conference news; Aug 2019 news). A later study corroborated the relationship, but the mechanism remained murky (Dec 2021 news).

To dissect how MS4A4A variants operate in the brain, first author Shih-Feng You used bulk RNA-Seq data from postmortem human cortical samples. The 579 donors came from three cohorts: ROSMAP, the Mayo RNA-Seq Study, and the Knight ADRC in St. Louis. They had died in their mid-80s, on average. Four hundred of them had been diagnosed with AD, the rest were age-matched controls.

You and colleagues first focused on the protective SNP rs1582763, which resides near the MS4A4A gene. A little more than one-third of donors carried this common allele. In bulk analysis, the variant altered expression of 1,071 microglial genes. Overall, it squelched expression of pro-inflammatory responses, while boosting anti-inflammatory interferon signaling and greasing lipid metabolism and cholesterol efflux. The inclusion of AD brains in the bulk sample likely amplified the number of activated microglia, and activation states, but the authors did not try to parse differences between AD and control brains.

The authors found the opposite pattern when they examined SNP rs6591561, a variant in the MS4A4A coding region that raises AD risk and was also present in nearly one-third of the cohort. This risk variant altered expression of 1,597 microglial genes, powering pro-inflammatory proteins such as cytokines, while suppressing interferon signaling and lipid metabolism genes. It lowered expression of MS4A4A itself. Some of these data were previously presented at the 2021 Alzheimer’s Association International Conference (Aug 2021 conference news). 

More Greens, Please. In people with the more common GG allele at the rs1582763 locus, few microglia assume the anti-inflammatory interferon profile (green). Carrying one or two copies of the protective A allele greatly expands this population. [Courtesy of You et al., medRxiv.]

How did the variants produce these opposing effects? Single-nuclei RNA-Seq data from 17,089 cortical microglia offered a clue. Isolated from 51 AD, seven control, and eight brains with other neurological conditions, the microglia fell into nine clusters. Cluster 3, characterized by expression of the chemokine receptor CXCR4, was most affected by the protective and risk SNPs. These microglia made more MS4A4A, and more interferons, than did microglia in other clusters. The protective variant boosted their MS4A4A expression and expanded this cluster, while the risk variant suppressed MS4A4A, dampened interferon signaling, and depleted these cells.

In particular, the protective variant seemed to push microglia from a chemokine-based expression profile toward a more interferon-based response. To test this idea, the authors expressed MS4A4A in human peripheral blood mononuclear cells. As in microglia, chemokine and inflammatory gene expression waned, while sTREM2 in the culture medium rose.

Altogether, the data suggested that MS4A4A helps microglia transition from a chemokine to an interferon state. “MS4A4A acts as a switch, and the risk variant represses that switch,” Karch noted. A recent study of human microglia in chimeric mice also identified distinct chemokine and interferon profiles (Mancuso et al., 2022).

How might cluster 3 arise? Statistical analysis of the snRNA-Seq data suggested that, under disease conditions, homeostatic microglia transition toward either the cluster 3 profile or a classic disease-associated microglia profile (Jun 2017 news). Cluster 3 microglia may represent a protective, anti-inflammatory state, the authors suggested. This jibes with an earlier mouse study that identified CXCR4-positive DAM microglia with an anti-inflammatory signature; these microglia efficiently cleared lipids and were specialized for phagocytosis (Rangaraju et al., 2018). 

You and colleagues wondered whether their findings suggest therapeutic approaches. Using a published Connectivity Map that details how small molecules affect gene expression, they identified 46 compounds predicted to mimic the effects of the protective variant, and 18 predicted to reverse the effects of the risk variant (Lamb et al., 2006; Subramanian et al., 2017). The former included several HDAC inhibitors, the latter, KIT and PDGFR inhibitors. “We can now test these drugs in animal models,” Karch noted. “This gives us a pathway to harness our findings therapeutically.”

Cruchaga believes the methods in this paper will be broadly applicable. “This is a good example of how informative and powerful snRNA-Seq is,” he wrote. “Many other groups are doing similar studies, which will be instrumental to understand the biology of Alzheimer’s GWAS hits.” Although not a co-author, Cruchaga collaborated on some aspects of this project.—Madolyn Bowman Rogers


  1. This manuscript provides a new insight to the mechanism underlying the contribution to Alzheimer’s disease of the MS4A locus, which distinguishes between protective and destructive microglia. It also highlights the connection between cholesterol metabolism and inflammation, a known phenomenon in other cells and organs, but novel in the context of microglia.

    Overall, the study emphasizes once again the complexity of microglial behavior under pathological conditions, and accordingly, the difficulty of successfully targeting microglia via a single therapeutic agent, in an attempt to downregulate their detrimental effects, while increasing the beneficial ones.

  2. You et al. used postmortem AD brain single nuclear RNA-Seq and identified one unique microglia signature that is functionally bidirectional by GWAS variants: an “interferon” cluster found in protective MS4A variant carriers, and a “chemokine” cluster in risk variant carriers. DAM identified from AD mouse brain harbor both anti- and pro-inflammatory profiles, raising controversy on microglial function for therapeutics. This study now determines that interferon is a signature that is protective, induced by enhanced MS4A4A expression, and connects nicely with functional downstream pathways, especially cholesterol efflux.

    It's been known that intracellular lipid accumulation in glia is a hallmark of AD. Which one is first—lipid or inflammation—is a chicken-and-egg question. The answer could be case-by-case, based on different variant-associated genes. For example, APOE is a lipid transporter, and APOE4 glia can display accumulated intracellular lipids, potentially inducing chemokines and proinflammatory factors (TCW et al., 2022). However, it is well appreciated that lipid and inflammation is tightly connected and co-regulated (van Diepen et al., 2013). This means we could potentially reverse the disease phenotype by modulating one or the other, as shown in this study demonstrating that the interferon signature increases cholesterol efflux.

    More work needs to be done to fully understand the disease mechanism. The approach the authors used—snRNA-Seq to identify a functional subcluster, especially well-defined gene sets connecting with functional downstream pathways—could unravel the mechanism of action of GWAS hits. Further, drug repurposing could be performed by testing whether drugs increase a given subpopulation of protective microglia (e.g., interferon).


    . Cholesterol and matrisome pathways dysregulated in astrocytes and microglia. Cell. 2022 Jun 23;185(13):2213-2233.e25. PubMed. BioRxiv.

    . Interactions between inflammation and lipid metabolism: relevance for efficacy of anti-inflammatory drugs in the treatment of atherosclerosis. Atherosclerosis. 2013 Jun;228(2):306-15. Epub 2013 Mar 1 PubMed.

  3. Genetic variants in the MS4A gene region were found to be associated with AD risk around 2011 by Naj et al. However, no functional follow-up studies have been done in this region because of our lack of understanding of the basic biology of this gene region. In 2019, my lab found that the same genetic variants in this region are the major regulators of sTREM2, and that at least the MS4A4A gene was one of the functional genes in this region (there may be more than one functional gene). This was important, as we provided a biological context to the original MS4A-AD signal: MS4A4A modifies risk for AD by modifying TREM2 biology, and it is involved in microglia fitness. However, the mechanism by which MS4A4A affects TREM2 biology is unclear.

    To understand the role of MS4A4A in microglia biology in the context of AD, the Karch lab used a very interesting and powerful study design, focusing on both the protective and risk MS4A4A variants that were found in the sTREM2 GWAS study by Deming et al., 2019. The Karch lab generated and analyzed brain single nuclei data from very well-characterized samples. By doing this, they found that specific MS4A4A alleles are associated with specific microglia populations.

    I think these results highlight just how dynamic microglia are, with many states, and different states may play different roles in disease. At the same time, this study opens the door to more questions. The authors clearly demonstrated that the protective alleles are associated with the microglia cluster with a lower inflammatory response, but it is not clear if MS4A4A is just a “passenger” in that cell cluster, or if it is really driving the microglia to that state, or how this links back to changes in sTREM2 levels. I expect additional studies from this group.

    In general, this is a good example from Karch lab of how informative and powerful single nuclei brain RNA-Seq is. Many other groups are doing similar studies, which will be, in fact, instrumental to understand the biology of the AD GWAS genes.


    . The MS4A gene cluster is a key modulator of soluble TREM2 and Alzheimer's disease risk. Sci Transl Med. 2019 Aug 14;11(505) PubMed.

  4. This is very nice, thought-provoking work. It shows the full potential of mining existing datasets to shed light on disease mechanisms. As a biologist, I highly appreciate the work from Shih-Feng You and collaborators because it gives us much more concrete paths to investigate in the lab, and to explore or prioritize specific pathways and targets.

    From a microglia point of view, the evidence here supports the idea that microglia are heterogenous and, whereas certain phenotypic states can be beneficial, others can drive or exacerbate disease. The next step would be to investigate what is the actual functional output of the MS4A-driven (and other) cell states, and to find ways to manipulate them in a desired direction.

    From a broader perspective, I think this is a fantastic way to make the bridge between genetic findings and cell function, and I am very eager to see what are the next stories coming from the Harari and Karch labs.

  5. Recent genome-wide association studies have revealed several AD-associated rare and common variants in genes that are selectively or preferentially expressed in microglia, including in the MS4A locus. Interestingly, SNPs rs1582763 (a protective variant) and rs6591561 (a risk variant) in this locus regulate the levels of soluble TREM2 (sTREM2) in the cerebrospinal fluid (CSF) and affect the age of onset in AD patients. More specifically, the protective rs1582763 variant has been previously shown to increase the levels of sTREM2 and delay the age of onset in AD patients, while the opposite effects have been observed with the AD risk-increasing rs6591561 variant.

    Given this background, You et al. have now assessed the transcriptomics signatures in human brain tissue samples harboring the above-mentioned MS4A locus variants by applying bulk RNA-Seq and single nuclei RNA-Seq (snRNA-Seq) approaches. Consequently, the authors show that the protective rs1582763 variant is linked to the increased expression of MS4A4A and to the shift of a chemokine microglia subpopulation toward an interferon state. Interestingly, opposite effects were observed in these same processes with the risk-increasing rs6591561 variant, suggesting that the expression state of MS4A4A is directly linked to the inflammatory response in microglia.

    As a further support for this idea, carrying the minor allele of the protective rs1582763 variant correlated with the appearance of a specific Mic.3 microglia cluster, which is an MS4A4A-dominat cluster persisting in an anti-inflammatory disease-associated microglia (DAM) state.

    Beyond this, the authors revealed that the expression of MS4A4A was connected to cholesterol and lipid metabolism in microglia as, e.g., CH25H encoding cholesterol 25-hydroxylase was found to be downregulated among the protective rs1582763 variant carriers. The authors further showed that the downregulation of CH25H in the cultured microglia led to a significant increase in the levels of sTREM2.

    The results now obtained by You et al. in relation to MS4A variants are in line with the recent functional discoveries in microglia, which have repeatedly pointed toward cholesterol and lipid dysfunction and altered interferon responses.

    Importantly, these above-mentioned results highlight much-needed mechanistic insights for the AD-associated variants in the MS4A locus. Thus, similar kinds of human tissue-derived approaches related to the other recently associated AD variants are needed to delineate biologically relevant disease mechanisms that may help to develop therapeutic strategies for AD in a personalized fashion.

  6. You et al. identified the bidirectional functions of MS4A variants, consistent with the “double-edged sword” hypothesis of innate immune activation and neurodegeneration. By single nucleus sequencing, this study demonstrated a detrimental MS4A variant relating to increased inflammasome response and intracellular cholesterol accumulation. It also showed a protective MS4A4A variant mediating reduced inflammatory gene expression and enhanced cholesterol metabolism, which shifted the pro-inflammatory microglia sub-cluster towards an interferon state.

    Previous studies showed that type-I interferon signaling is detrimental in AD mice, which is mediated by enhanced neuroinflammation, synaptic loss, and protein aggregation (Roy et al., 2020; 2022).

    On the other hand, other research showed a protective role of type-I interferon in acute settings (Chavoshinezhad et al., 2019). Moreover, Aβ-specific Th1 cells secreting IFNg induced the expression of MHCII and enhanced Aβ-plaque clearance in 5XFAD mice (Mittal et al., 2019). Here, this study emphasized the beneficial role of interferon in AD, an emerging mediator for neurodegeneration.

    This study is a good example of the complexity of GWAS risk loci and their functions in diseases. It attempts to link the genetic findings, biological functions, and potential drug candidates, which could be applied to study the roles of other LOAD GWAS risk factors.


    . Type I interferon response drives neuroinflammation and synapse loss in Alzheimer disease. J Clin Invest. 2020 Apr 1;130(4):1912-1930. PubMed.

    . Concerted type I interferon signaling in microglia and neural cells promotes memory impairment associated with amyloid β plaques. Immunity. 2022 May 10;55(5):879-894.e6. Epub 2022 Apr 19 PubMed.

    . Intranasal interferon beta improves memory and modulates inflammatory responses in a mutant APP-overexpressing rat model of Alzheimer's disease. Brain Res Bull. 2019 Aug;150:297-306. Epub 2019 Jun 21 PubMed.

    . CD4 T Cells Induce A Subset of MHCII-Expressing Microglia that Attenuates Alzheimer Pathology. iScience. 2019 Jun 28;16:298-311. Epub 2019 May 30 PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Large Genetic Analysis Pays Off With New AD Risk Genes
  2. MS4A Alzheimer’s Risk Gene Linked to TREM2 Signaling
  3. Parsing How Alzheimer’s Genetic Risk Works Through Microglia
  4. Paper Alert: MS4A Variants May Sway Alzheimer’s Risk Via TREM2
  5. Massive Proteomics Study Connects Genes, Proteins, Disease
  6. ADAD and LOAD: At Cellular Level, They Are Not the Same
  7. Hot DAM: Specific Microglia Engulf Plaques

Paper Citations

  1. . A multi-pronged human microglia response to Alzheimer’s disease Aβ pathology. bioRXiv. 7 Jul 2022 bioRxiv
  2. . Identification and therapeutic modulation of a pro-inflammatory subset of disease-associated-microglia in Alzheimer's disease. Mol Neurodegener. 2018 May 21;13(1):24. PubMed.
  3. . The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science. 2006 Sep 29;313(5795):1929-35. PubMed.
  4. . A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles. Cell. 2017 Nov 30;171(6):1437-1452.e17. PubMed.

Further Reading

Primary Papers

  1. . MS4A4A modifies the risk of Alzheimer disease by regulating lipid metabolism and immune response in a unique microglia state. medRxiv. 2023 Feb 8; PubMed.